fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2025-12-30 14:20:11 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

aco: implement subgroup shader_clock on GFX10.3

Browse source 
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5546>
This commit is contained in:
Rhys Perry 2020-06-18 14:45:31 +01:00
parent a5303a3cbe
commit 07250a92da
3 changed files with 16 additions and 6 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

17
src/amd/compiler/aco_instruction_selection.cpp
View file

@ -8155,11 +8155,18 @@ void visit_intrinsic(isel_context *ctx, nir_intrinsic_instr *instr)
break;
}
case nir_intrinsic_shader_clock: {
aco_opcode opcode =
nir_intrinsic_memory_scope(instr) == NIR_SCOPE_DEVICE ?
aco_opcode::s_memrealtime : aco_opcode::s_memtime;
bld.smem(opcode, Definition(get_ssa_temp(ctx, &instr->dest.ssa)), memory_sync_info(0, semantic_volatile));
emit_split_vector(ctx, get_ssa_temp(ctx, &instr->dest.ssa), 2);
Temp dst = get_ssa_temp(ctx, &instr->dest.ssa);
if (nir_intrinsic_memory_scope(instr) == NIR_SCOPE_SUBGROUP && ctx->options->chip_class >= GFX10_3) {
/* "((size - 1) << 11) | register" (SHADER_CYCLES is encoded as register 29) */
Temp clock = bld.sopk(aco_opcode::s_getreg_b32, bld.def(s1), ((20 - 1) << 11) | 29);
bld.pseudo(aco_opcode::p_create_vector, Definition(dst), clock, Operand(0u));
} else {
aco_opcode opcode =
nir_intrinsic_memory_scope(instr) == NIR_SCOPE_DEVICE ?
aco_opcode::s_memrealtime : aco_opcode::s_memtime;
bld.smem(opcode, Definition(dst), memory_sync_info(0, semantic_volatile));
}
emit_split_vector(ctx, dst, 2);
break;
}
case nir_intrinsic_load_vertex_id_zero_base: {

2
src/amd/compiler/aco_opt_value_numbering.cpp
View file

@ -219,6 +219,8 @@ struct InstrPred {
switch (a->format) {
case Format::SOPK: {
if (a->opcode == aco_opcode::s_getreg_b32)
return false;
SOPK_instruction* aK = static_cast<SOPK_instruction*>(a);
SOPK_instruction* bK = static_cast<SOPK_instruction*>(b);
return aK->imm == bK->imm;

3
src/amd/compiler/aco_scheduler.cpp
View file

@ -459,7 +459,8 @@ HazardResult perform_hazard_query(hazard_query *query, Instruction *instr, bool
/* don't move non-reorderable instructions */
if (instr->opcode == aco_opcode::s_memtime ||
instr->opcode == aco_opcode::s_memrealtime ||
instr->opcode == aco_opcode::s_setprio)
instr->opcode == aco_opcode::s_setprio ||
instr->opcode == aco_opcode::s_getreg_b32)
return hazard_fail_unreorderable;
memory_event_set instr_set;